Research On Mobile Robot Vision SLAM Algorithm For Dynamic Environment

With the development of information technology and artificial intelligence,autonomous mobile robot has gradually become an important research direction and research focus in the field of robotics.The main technical problems of mobile robot include localization,mapping,real-time obstacle avoidance and path planning.Localization and mapping is a hot research topic in the field of autonomous mobile robot.In this paper,we research on visual SLAM(simultaneous localization and mapping)algorithm in order to improve its robustness and accuracy in dynamic scenes,focus on lightweight visual SLAM algorithm design,improvement of the accuracy of map points and visual-inertial SLAM algorithm design.The main work of this paper is as follows:1.Visual SLAM algorithm lightweight method design.The use of NEON instruction for feature extraction and description part reduces the running time on ARM.The center point method was put forward,a new algorithm for local mapping,has reduced time consumption of the projection matching.We also divided feature extraction,feature description and pose estimation into different thread,through the parallel work of the three parts of the front speed algorithm.2.Map points management strategy design.The error model of the binocular camera is set up.Based on the analysis of the source of the distance error of the binocular camera,a map initialization algorithm is proposed to provide a guarantee for the localization and mapping accuracy.The feature points management algorithm was proposed to improve the accuracy of visual SLAM algorithm at some extent.This algorithm used depth filter at front-end for fusing point-depth.Also it was able to feature points in mapping,which have matching formation within continuous graphs and with uncertain depth convergent.3.Research on Visual-inertial SLAM algorithm.This study introduces IMU(Inertial measurement unit)initialization algorithm based on binocular cameras,to guarantee the correct integration of IMU.By using IMU integral model instead of constant velocity model as the initial motion estimation between adjacent frames,robustness of the SLAM algorithm was improved.4.Experimental research on visual SLAM algorithm for dynamic environment.The sensor system layout and mobile robot platform are built,and the accuracy and robustness of the SLAM algorithm is verified.The experimental results verify the feasibility and effectiveness of the algorithm.